Regional Brain Volume Differences Associated with Hyperglycemia and Severe Hypoglycemia in Youth with Type 1 Diabetes

Pathophysiology/Complications ORIGINAL ARTICLE

Regional Brain Volume Differences Associated With Hyperglycemia and Severe Hypoglycemia in Youth With Type 1 Diabetes

1 1,3,4,5 DANA C. PERANTIE, BS KEVIN J. BLACK, MD cemia during childhood with deficits in 1 6 JENNY WU, BA MICHELLE SADLER, RN, BSN, CDE specific cognitive domains (3,4). These 1 6,7 JONATHAN M. KOLLER, BS NEIL H. WHITE, MD, CDE findings suggest that during develop- 1 1,3,4 AUDREY LIM, BA TAMARA HERSHEY, PHD 2 ment, exposure to glycemic extremes may STACIE L. WARREN, BA alter the structure or function of specific pathways or regions in the brain. Recent brain imaging studies in diabetic adults OBJECTIVE — Despite interest in the effects of type 1 diabetes on the developing brain, have reported differences in grey or white structural brain volumes in youth with this disease have not previously been examined. This matter integrity associated with prior study is the first to quantify regional brain volume differences in a large sample of youth with hypo- or hyperglycemia (5,6). However, diabetes. the effects of diabetes on the developing RESEARCH DESIGN AND METHODS — Magnetic resonance images (MRIs) were brain have not been assessed in any large- acquired from youth with diabetes (n ϭ 108) and healthy sibling control subjects (n ϭ 51) aged scale study to date (7). Assessing brain 7–17 years. History of severe hypoglycemia was assessed by parent interview and included integrity earlier in the course of brain de- seizure, loss of consciousness, or requiring assistance to treat. A1C values since diagnosis were velopment and diabetes, followed by pro- obtained from medical records; median A1C was weighted by duration of disease. Voxel-based spective monitoring, would be essential morphometry was used to determine the relationships of prior hypo- and hyperglycemia to to determine when differences may regional grey and white matter volumes across the whole brain. emerge. Such knowledge could shed light on the neural basis of observed cognitive RESULTS — No significant differences were found between diabetic and healthy control effects in children and adults with diabe- groups in grey or white matter. However, within the diabetic group, a history of severe hypo- tes and determine whether there are de- glycemia was associated with smaller grey matter volume in the left superior temporal region. Greater exposure to hyperglycemia was associated with smaller grey matter volume in the right velopmental time periods during which cuneus and precuneus, smaller white matter volume in a right posterior parietal region, and the brain may be particularly vulnerable larger grey matter volume in a right prefrontal region. to the negative effects of hypoglycemia or hyperglycemia. CONCLUSIONS — Qualitatively different relationships were found between hypo- and The present study is the first to exam- hyperglycemia and regional brain volumes in youth with type 1 diabetes. Future studies should ine the structural integrity of the brain in investigate whether these differences relate to cognitive function and how these regions are a large sample of children and adolescents affected by further exposure. with type 1 diabetes. We used high- resolution structural magnetic resonance Diabetes Care 30:2331–2337, 2007 imaging (MRI) and voxel-based morphometry (VBM), an objective method of ype 1 diabetes is known to have cu- The effects of diabetes on central nervous quantitatively analyzing MRI data, to de- mulative deleterious effects on the system structure and function are less termine whether exposure to hypo- or hy- T body, most notably on the retina, well understood. A number of studies as- perglycemia in youth with type 1 diabetes kidney, nerves, and blood vessels (1,2). sociate exposure to hypo- and hypergly- is associated with differences in grey or ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● white matter volumes. From the 1Department of Psychiatry, Washington University School of Medicine, St. Louis, Missouri; the 2Department of Neurology, Washington University School of Medicine, St. Louis, Missouri; the 3Department RESEARCH DESIGN AND of Radiology, Washington University School of Medicine, St. Louis, Missouri; the 4Department of Pediatrics, METHODS 5 — Children aged 7–17 Washington University School of Medicine, St. Louis, Missouri; the Department of Anatomy and Neuro- years with type 1 diabetes and nondia- biology, Washington University School of Medicine, St. Louis, Missouri; 6St. Louis Children’s Hospital, St. Louis, Missouri; and the 7Department of Psychology, University of Illinois, Urbana-Champaign, Illinois. betic siblings (healthy control subjects) Address correspondence and reprint requests to Tamara Hershey, PhD, Campus Box 8225, 4525 Scott were recruited from the Diabetes Clinic at Ave., Washington University School of Medicine, St. Louis, MO 63110. E-mail: [email protected]. St. Louis Children’s Hospital, which is af- Received for publication 19 February 2007 and accepted in revised form 7 June 2007. filiated with Washington University in St. Published ahead of print at http://care.diabetesjournals.org on 15 June 2007. DOI: 10.2337/dc07-0351. Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/ Louis. Subjects were excluded for mental dc07-0351. retardation, chronic disease other than Abbreviations: MRI, magnetic resonance imaging; VBM, voxel-based morphometry. type 1 diabetes (e.g., hypothyroidism), A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion significant neurological history not due to factors for many substances. diabetes, diagnosed psychiatric disorder, © 2007 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby current use of psychoactive medications, marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. prematurity at birth Ͼ4 weeks early with

DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 2331 Regional brain volume and type 1 diabetes complications, and contraindications to aging system with a standard Siemens are not attributable to global differences MRI (e.g., metal implants). To reduce the 30-cm circularly polarized RF head coil. in volume, such as those expected be- likelihood of residual ␤-cell function, di- For each subject, three to five images con- tween sexes and with age (16). Age, sex, abetic subjects were required to have been sisting of 128 contiguous 1.25-mm sagittal and total volume of the relevant brain tis- diagnosed and on insulin for at least 2 slices were acquired using magnetization- sue (grey/white) were removed as covari- years. Handedness was assessed with a prepared rapid gradient echo. Subjects ates from all models. In addition, in modified Edinburgh Handedness Inven- with movement or other artifact were ex- models using only diabetic subjects, age of tory (8). Procedures were approved by cluded (n ϭ 10). Images with suspected onset was also covaried. Independent sam- the Washington University School of anatomical abnormalities were referred to ple t tests were performed for compari- Medicine’s Human Studies Committee, a neuroradiologist for review; three sub- sons between groups, defining contrasts and all participants and their parents or jects were excluded for confirmed brain in each direction (e.g., any hypoglyce- guardians signed informed consents. abnormalities. For each subject, three mia Ͼ no hypoglycemia; any hypoglyce- high-quality images were averaged after mia Ͻ no hypoglycemia). For analyses of Clinical variables being coregistered by an automated, vali- hyperglycemia exposure, multiple regres- Detailed information about each diabetic dated technique (10). sions were performed with contrasts in youth’s history of severe hypoglycemia, each direction (negative and positive hyperglycemia, and other diabetes com- Image analysis correlations). Cluster level multiple complications was collected by parental and VBM was performed with statistical para- parisons–corrected P values Ͻ0.05 were child interview. Severe hypoglycemia was metric mapping software (SPM5; Well- considered significant. defined as events with neurological dys- come Department of Cognitive Neurology function, including seizure, loss of con- [available at www.fil.ion.ucl.ac.uk]). Im- RESULTS — A total of 108 youth with sciousness, or inability to arouse from ages were simultaneously normalized to type 1 diabetes and 51 healthy control sleep, or those requiring assistance of Montreal Neurological Institute space, subjects were included in these analyses. someone other than the patient for treat- corrected for intensity inhomogeneity, See Table 1 for demographic and clinical ment (9). Hyperglycemic history was es- and tissue segmented as grey matter, information and Table 2 for a summary of timated from all available A1C test results, white matter, and cerebrospinal fluid imaging results. collected from participants’ medical based on a priori probability maps (11). records at St. Louis Children’s Hospital. Grey and white segments were modulated Type 1 diabetic versus healthy A1C tests approximate blood glucose to produce images representing grey and control subjects control over the previous 2–3 months. white matter volume (11). After this pro- Diabetic and healthy control groups did The amount of time represented by the cessing, voxel dimensions were 2 ϫ 2 ϫ 2 not differ significantly in sex distribution A1C tests was calculated by multiplying mm. Modulated segments were smoothed (␹2 ϭ 0.58, P ϭ 0.45), mean age (t ϭ the number of tests by 3 months and di- with a 12-mm full-width at half- Ϫ0.59, P ϭ 0.56), or mean parental edu- viding by duration of diabetes in months. maximum Gaussian kernel to promote cation (t ϭ 0.42, P ϭ 0.68) (Table 1). The Participants with A1C coverage for Ͻ30% normality of residuals (12). diabetic group had proportionally more of their duration of diabetes (n ϭ 10) were Voxels with segmented intensities left-handed or ambidextrous subjects excluded from hyperglycemia analyses. Ͻ0.1 were masked out with an absolute than the healthy control subjects (␹2 ϭ Less-than-complete coverage was due to threshold to reduce voxels possibly be- 3.86, P ϭ 0.05) (Table 1). In VBM analy- clinical appointments Ͼ3 months apart, longing to other tissue classes and be- ses comparing diabetic versus healthy transfers from other clinics, or use of total cause these voxels are less likely to adhere control groups, there were no significant glycated hemoglobin tests. To account for to assumptions of normality (13). Images grey or white matter volume differences. duration of exposure to hyperglycemia, a were analyzed by SPM5, performing stan- Covarying handedness did not change “hyperglycemia exposure score” was cal- dard parametric tests (e.g., regressions) at these results. culated. Because a child with an average each voxel, which results in statistical A1C of 8% and duration of diabetes for 10 parametric maps on which every voxel’s Hypoglycemia years has had more exposure to hypergly- intensity corresponds to a t value. The sta- Because the distribution of severe hypo- cemia than has a child with the same av- tistical parametric maps were then thresh- glycemic episodes was skewed, with most erage A1C and duration of diabetes for 2 olded to show only voxels with t values subjects having few or no episodes (me- years, a score that weighted duration and corresponding to uncorrected P Ͻ 0.001. dian 1 episode [range 0–50]; lower quar- A1C equally was calculated by adding The probability of resulting clusters was tile ϭ 0, median quartile ϭ 1, upper each patient’s z score of median A1C to corrected for multiple comparisons using quartile ϭ 2), subjects were categorized the z score of duration of diabetes. This the stat_threshold script from Worsley’s as having no (n ϭ 42) or any (n ϭ 66) method of calculation results in a near- fmristat package (14). This cluster level severe hypoglycemic episodes. Age, normal distribution of scores, with higher method of multiple comparisons correc- handedness, sex, parental education, es- scores indicating more exposure to hyper- tion takes into account nonuniformity timated IQ, and median A1C did not dif- glycemia. Each child’s hyperglycemia ex- due to intrinsically inhomogeneous fer between the no (n ϭ 42) and any (n ϭ posure score can be interpreted relative to smoothness (15). 66) hypoglycemia groups, but the any hy- this sample only. Total volume of each tissue class (grey poglycemia group had longer duration of matter or white matter) was calculated by diabetes (t ϭϪ5.03, P Ͻ 0.001), earlier Image acquisition summing modulated voxel intensities for age of onset (t ϭ 3.49, P ϭ 0.001), and Structural images were acquired for each that class. Covarying total grey/white higher hyperglycemia exposure scores subject on a Siemens Sonata 1.5 Tesla im- matter volume ensures that differences (t ϭϪ4.73, P Ͻ 0.001) (Table 1).

2332 DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 Perantie and Associates

Table 1—Demographic and clinical variables

Control subjects Diabetic subjects No hypoglycemia Any hypoglycemia n 51 108 42 66 Age 12.3 Ϯ 2.7 12.6 Ϯ 2.7 12.3 Ϯ 2.4 12.8 Ϯ 2.9 Sex ͓male/female (% male)͔ 26/25 (52) 62/46 (57) 27/15 (64) 35/31 (53) Race ͓white/minority (% minority)͔* 43/7 (14) 95/12 (11) 37/7 (16) 60/5 (8) Handedness ͓right-handed/other (% right-handed)͔ 50/1 (98) 96/12 (89)† 37/5 (88) 59/7 (89) Parental education‡ 15.2 Ϯ 2.2 15.0 Ϯ 2.2 14.8 Ϯ 2.5 15.2 Ϯ 2.0 Duration of diabetes 5.7 Ϯ 2.9 4.1 Ϯ 1.9 6.7 Ϯ 3.0§ Age of onset 6.9 Ϯ 3.3 8.2 Ϯ 2.9 6.1 Ϯ 3.2§ Median A1C¶ 8.4 Ϯ 1.0 8.2 Ϯ 1.1 8.4 Ϯ 0.8 Hyperglycemia exposure score¶ 0.0 Ϯ 1.4 Ϫ0.8 Ϯ 1.2 0.5 Ϯ 1.3§ Data are means Ϯ SD unless otherwise indicated. *Not reported for two participants. †Significantly different from healthy control subjects; P Ͻ 0.05. ‡Not reported for five participants; §Any hypoglycemia significantly different from no hypoglycemia; P Ͻ 0.001. ¶Ten subjects with Ͻ30% coverage excluded.

In VBM analyses, the any hypoglyce- Age of onset served (19). This region has been associ- mia group had less grey matter volume Given the possibility that age of onset ated with the episodic memory system than that in the no hypoglycemia group in might confound our results (despite hav- (20); severe hypoglycemia has previously the left superior temporal/occipital cortex ing covaried it from those analyses), we been found to affect episodic memory in (P ϭ 0.001) and left inferior occipital cor- performed an exploratory analysis corre- children (3,21), but the relation of such tex (P ϭ 0.0002) (Fig. 1A). Because the lating age of onset to grey and white mat- cognitive changes to the current brain any hypoglycemia group also had higher ter volume. These analyses found that findings remains to be examined. This hyperglycemia exposure scores, we addi- earlier age of onset was related to larger area is also part of the “default system,” a tionally covaried hyperglycemia exposure white matter volume in the left precuneus set of interconnected brain regions with scores. Notably, the volume of the left region (voxel extent ϭ 177; P ϭ 0.02). high resting state neuronal activity that temporal/occipital region was still smaller There were no significant differences for decreases in response to cognitive chal- (P ϭ 0.008), but the left inferior occipital white matter in the other direction or for lenges (22). It is possible that this high cortex region was not (P ϭ 0.13). There grey matter in either direction. baseline rate of blood flow creates a were no differences in grey matter in the heightened vulnerability of the neurons other direction (any hypoglycemia Ͼ no CONCLUSIONS — This is the first to significantly reduced blood glucose hypoglycemia) and no differences in study to examine the effects of type 1 di- during hypoglycemia. white matter volume in either direction. abetes on brain structure in a large sample While no direct comparisons were No differences were found comparing the of children and adolescents. We found made between the left and right hemi- healthy control group to the any or the no that youth with type 1 diabetes did not spheres, our data may suggest a stronger hypoglycemia groups. significantly differ from nondiabetic sib- effect of severe hypoglycemia on the left lings in regional grey or white matter vol- side of the brain than that on the right Hyperglycemia umes. However, within the diabetic side. At least three neuropathological case The mean Ϯ SD number of A1C values group, we found qualitatively different re- studies also described more extensive per subject was 14.3 Ϯ 6.6; the percent- lationships between exposure to severe damage in the left than in the right hemi- age of the duration of diabetes repre- hypoglycemia and chronic hyperglyce- sphere with severe hypoglycemia sented by A1C tests was 68 Ϯ 14%. mia and regional grey and white matter (18,23,24). Additionally, in a single pho- Hyperglycemia exposure scores were nor- volumes. These differences were statisti- ton emission computed tomography mally distributed (0.00 Ϯ 1.42 [range cally significant despite the relatively (SPECT) study, children with diabetes Ϫ2.45 to 3.69]) and treated as a continu- short duration of diabetes in our sample. had abnormal left-right blood flow ratios, ous variable. In VBM analyses, higher hy- suggestive of left hemisphere dysfunction perglycemia exposure scores correlated Hypoglycemia (25). Interestingly, in the default system, with less grey matter volume in the right Compared with their hypoglycemia-naive this same region has a greater response to cuneus and precuneus (P ϭ 0.02) (Fig. diabetic peers, diabetic youth with one or cognitive challenge on the left than on the 1B). Hyperglycemia exposure scores also more prior severe hypoglycemic episodes right side (22). correlated with larger grey matter volume had smaller grey matter volume at the left in the right frontal middle gyrus (P ϭ temporal-occipital junction. Interest- Hyperglycemia 0.008) (Fig. 1C) and with smaller white ingly, smaller grey matter in a similar re- The extent of exposure to hyperglycemia matter volume in right superior parietal gion (left superior temporal and angular was associated with differences in both white matter (P ϭ 0.01) (Fig. 1D). A sim- gyri) has been reported in adults with grey and white matter volumes. Smaller ilar cluster appeared in the left superior type 1 diabetes (5). In addition, human grey matter volume was found in poste- parietal white matter but did not survive neuropathological case studies of pro- rior cortical areas (right cuneus and pre- multiple comparisons correction (P ϭ found hypoglycemia in adults have re- cuneus). This region is associated with 0.13). Higher hyperglycemia exposure ported defects in temporal and/or higher-order visuospatial function and scores were not associated with greater occipital regions (17,18); however, in one episodic memory (26); the relation of white matter volume in any region. report, these regions were relatively pre- findings in this region to diminished per-

DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 2333 Regional brain volume and type 1 diabetes

formance in these cognitive domains in t Peak voxel diabetic children (3,21) remains to be in- vestigated. A study in adults with type 1

12 4.65 diabetes also found less grey matter in the ؊ 24 4.39 value. AH, any

t right cuneus in those with higher lifetime Ϫ 72, 2084, 4.88 54, 40 4.22 66, 14 3.95 A1C averages (5). In addition, another Peak 54, 52 4.29 74, 22 4.68 ؊ ؊ Ϫ Ϫ (x, y, z) ؊ ؊ study reported lower grey matter density voxel MNI coordinates 22, 10, 58, 52, 42, 22, in the right occipital lobe in diabetic Ϫ Ϫ ؊ ؊ Ϫ adults with retinopathy compared with that in healthy control subjects (27). The

P mechanism for this regional effect is un-

cluster known, but Wessels et al. (27) speculated Corrected that the occipital lobe is at risk for hypo- perfusion because of its location in a “wa- tershed area” of circulation. One might size Cluster (voxels) expect these effects to correlate to vascular changes. Although vascular disease

otal white matter volume. and retinopathy were not assessed in this sample, some degree of early vascular

area(s) changes might be present in diabetic chil-

Brodmann’s dren within a few years of onset (1).

2 x 2 mm. Peak voxel refers to the voxel with the greatest Therefore, a vascular mechanism cannot

ϫ be ruled out as accounting for this finding even in this age range. This region, also in the “default system,” has been noted for having the highest baseline metabolism of the whole brain (22) and may be prefer- entially vulnerable to insults (28). L superior parietal Near 5, 7 178 0.13 L ventral prefrontal 11 163 0.18 L inferior occipital 19, 18 508 0.002 Our analyses also revealed greater grey matter volume in a right prefrontal region with greater exposure to hyperglycemia. Overall, grey matter decreases HCHC None None HC None HC L calcarine, precuneus 18, 17 287 0.14 with development across this age range Ͻ Ͼ Ͼ Ͻ (29), so this finding may reflect an ab- NH L temporal-occipital 39, 37, 19 901 0.001 NH None NH None NH None Contrast Region normal developmental trajectory. Alter- < Ͻ Ͼ Ͼ natively, this finding may indicate a Positive correlation None Diabetic AH Negative correlation R superior parietal Near 5, 7 297 0.01 22, AH Negative correlation R cuneus, precuneus 18, 19, 7 464 0.02 12, Positive correlation R prefrontal 45, 46, 9, 8 473 0.008 46, 34, 36 3.87 Diabetic AH compensatory reaction to the lower grey matter volume that we found in the right cuneus and precuneus. Musen et al. (5) n also reported greater grey matter density associated with higher lifetime A1C in adults with diabetes, but the affected region was in a different location (parietal 0.20 are shown; significant results are in bold. Voxel dimensions are 2 lobe) than the ones that we report here in Ͻ

P children. White matter volume was lower in the right superior parietal region in subjects with greater hyperglycemia exposure, with a similar finding in the homologous area on the left side. The significant region on the right was adjacent to the hyperglycemia-associated grey matter area (precuneus), and these two findings may be related. By reconstructing white matter fi- ber bundles and their neuroanatomical connectivity to grey matter regions, diffusion tensor tractography could further address whether these two regions may be Summary of models, contrasts, and results connected. A magnetic resonance spec- troscopy study in children with a history of significant hyperglycemia reported low White Diabetic vs. HC subjectsWhite AH vs. NH Age, sex, WM 108 vs. 51 Diabetic Age, sex, WM, onset 66 vs. 42 AH White Hyperglycemia exposure score Age, sex, WM, onset 98 Grey Hyperglycemia exposure score Age, sex, GM, onset 98 Tissue type Model Covariates Grey Any vs. NH Age, sex, GM, onset 66 vs. 42 Grey Diabetic vs. HC Age, sex, GM 108 vs. 51 Diabetic Table 2— Regions with multiple-comparison corrected cluster level hypoglycemia; GM, total grey matter volume; HC, healthy control; L, left; MNI, Montreal Neurological Institute; NH, no hypoglycemia; R, right; WM, t metabolite ratios in posterior parietal

2334 DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 Perantie and Associates

Figure 1—Results overlaid on individual subject’s brain in Montreal Neurological Institute space. A: Regions with smaller grey matter volume in diabetic youth with histories of severe hypoglycemia compared with those in diabetic youth without histories of severe hypoglycemia. Regions of less grey matter (B), more grey matter (C), and less white matter (D) associated with greater hyperglycemia exposure. Crosshairs indicate location of peak voxel of the significant region. The color bar indicates t values. In the coronal view, the left side of the images depicts the left side of the brain. A supplementary figure is available in the online appendix at http://dx.doi.org/10.2337/dc07-0351. white matter, indicating possible dys- The finding that age of onset was not as- from the volume difference in the precu- function or reduced axonal density in this sociated with differences in grey matter is neus region found to be associated with region (30). In normal children, anisot- consistent with a previous report, which hyperglycemia exposure (see above). ropy in the parietal white matter has been found no effect of diabetes-onset age on Thus, the anatomical location and the di- found to increase with age (31) and to traced temporal lobe or amygdalohippo- rection of the relationship suggest that age correlate with IQ (32). Thus, it is possible campal volumes in young adults with of onset does not explain the results from that altered white matter volume in this type 1 diabetes (33). The association of our analyses of the effects of hypoglyce- region could be reflected in aspects of earlier age of onset with larger white mat- mia and hyperglycemia. cognitive performance. ter volume near the left precuneus was As with complications in other organs unexpected, but there have been no pre- and systems in youth with type 1 diabetes Age of onset vious studies examining the effects of age (e.g., neuropathy, retinopathy, and ne- Age of onset of diabetes was examined as of onset on white matter volume. Our phropathy) (1,2), the effects of blood glu- a potential confounder to the observed ef- finding was in the opposite direction cose extremes on the structural integrity fects of hypoglycemia and hyperglycemia. (larger) and on the opposite side (left) of the brain, although significant, may be

DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 2335 Regional brain volume and type 1 diabetes subclinical at this early stage. These subtle 13. Ashburner J, Friston KJ: Voxel-based References morphometry: the methods. Neuroimage differences are observable at a group level 1. Olsen BS, Sjolie A, Hougaard P, Johan- and are not likely to be apparent in indi- 11:805–821, 2000 nesen J, Borch-Johnsen K, Marinelli K, 14. Worsley KJ, Liao CH, Aston J, Petre V, vidual patients at this age. However, these Thorsteinsson B, Pramming S, Mortensen differences were generated with conserva- Duncan GH, Morales F, Evans AC: A gen- HB, the Danish Study Group of Diabetes eral statistical analysis for fMRI data. Neu- tive statistical methods designed to mini- in Childhood: A 6-year nationwide cohort roimage 15:1–15, 2002 mize the probability of false positives study of glycaemic control in young peo- 15. Moorhead TW, Job DE, Spencer MD, (34), so their potential consequences ple with type 1 diabetes: risk markers for Whalley HC, Johnstone EC, Lawrie SM: should be examined. Currently, it is un- the development of retinopathy, ne- Empirical comparison of maximal voxel known whether these differences are as- phropathy and neuropathy. J Diabetes and non-isotropic adjusted cluster extent sociated with cognitive consequences. It Complications 14:295–300, 2000 results in a voxel-based morphometry is possible that, as with other complica- 2. Riihimaa PH, Suominen K, Tolonen U, study of comorbid learning disability with Jantti V, Knip M, Tapanainen P: Periph- tions, these brain differences and their schizophrenia. Neuroimage 28:544–552, eral nerve function is increasingly im- 2005 consequences could be compounded by paired during puberty in adolescents with further cumulative exposure to glycemic 16. Gogtay N, Giedd JN, Lusk L, Hayashi KM, type 1 diabetes. Diabetes Care 24:1087– Greenstein D, Vaituzis AC, Nugent TF, extremes with advancing age and increas- 1092, 2001 III, Herman DH, Clasen LS, Toga AW, ing duration of diabetes. 3. Hershey T, Perantie DC, Warren SL, Zim- Rapoport JL, Thompson PM: Dynamic We propose that the regional volume merman EC, Sadler M, White NH: Fre- mapping of human cortical development differences detected in this study reflect quency and timing of severe hypoglycemia during childhood through early adult- the impact of hyper- and hypoglycemia affects spatial memory in children with hood. Proc Natl Acad Sci U S A 101:8174– on neural integrity and/or development. type 1 diabetes. Diabetes Care 28:2372– 8179, 2004 In animal models, hypoglycemia has been 2377, 2005 17. Akyol A, Kiylioglu N, Bolukbasi O, Guney shown to induce neuronal death and dys- 4. Northam EA, Anderson PJ, Jacobs R, E, Yurekli Y: Repeated hypoglycemia and Hughes M, Warne GL, Werther GA: Neu- function (35), and hyperglycemia is re- cognitive decline: a case report. Neuroen- ropsychological profiles of children with docrinol Lett 24:54–56, 2003 ported to cause injury to myelin and type 1 diabetes 6 years after disease onset. neurons (36); these data support our sup- 18. Jones GM: Posthypoglycemic encepha- Diabetes Care 24:1541–1546, 2001 lopathy. Am J Med Sci 213:206–213, 1947 position that glycemic extremes precede 5. Musen G, Lyoo KL, Sparks C, Weinger K, 19. Lawrence RD, Meyer A, Nevin S: The measurable differences in the brain. How- Renshaw PF, Hwang J, Ryan CM, Jimer- pathological changes in the brain in fatal ever, because of the retrospective and cor- son DC, Jacobson AM: Evidence for re- hypoglycemia. Quart J Med 11:181–201, relative nature of this study, we cannot duced grey matter density in patients with 1942 rule out that the differences reported here type 1 diabetes as measured by magnetic 20. Cabeza R, Prince SE, Daselaar SM, Green- were present before exposure to glycemic resonance imaging (Abstract). Diabetes 52 berg DL, Budde M, Dolcos F, Labar KS, extremes or diabetes. Retrospective re- (Suppl. 1):A57–A58, 2004 Rubin DC: Brain activity during episodic port of severe glycemic experiences has 6. Wessels AM, Rombouts SA, Simsek S, retrieval of autobiographical and labora- Kuijer JP, Kostense PJ, Barkhof F, Schel- tory events: an fMRI study using a novel been found to be fairly reliable in adults tens P, Snoek FJ, Heine RJ: Microvascular (37), but prospective measures are likely photo paradigm. J Cogn Neurosci 16: disease in type 1 diabetes alters brain ac- 1583–1594, 2004 to be more accurate, especially over long tivation: a functional magnetic resonance 21. Northam EA, Anderson PJ, Werther GA, time periods (38). Prospective follow-up imaging study. Diabetes 55:334–340, Warne GL, Andrewes D: Predictors of of our sample is ongoing and should be 2006 change in the neuropsychological profiles able to determine if further exposure to 7. Northam EA, Rankins D, Cameron FJ: of children with type 1 diabetes 2 years hyper- and hypoglycemia accentuates the Therapy insight: the impact of type 1 di- afterdiseaseonset.DiabetesCare22:1438– pattern of regional volume differences re- abetes on brain development and func- 1444, 1999 ported here. It should also be noted that tion. Nat Clin Pract Neurol 2:78–86, 2006 22. Raichle ME, MacLeod AM, Snyder AZ, glycemic extremes could affect the func- 8. Oldfield RC: The assessment and analysis Powers WJ, Gusnard DA, Shulman GL: A tion of the brain with or without altering of handedness: the Edinburgh inventory. default mode of brain function. Proc Natl Neuropsychologia 9:97–113, 1971 regional volumes. Future studies using Acad SciUSA98:676–682, 2001 9. The DCCT Research Group: Epidemiol- 23. Chalmers J, Risk MTA, Kean DM, Grant R, methods such as functional MRI would be ogy of severe hypoglycemia in the diabe- Ashworth B, Campbell IW: Severe amne- needed to determine whether functional tes control and complications trial. Am J sia after hypoglycemia: clinical, psycho- abnormalities exist in this population. Med 90:450–459, 1991 metric, and magnetic resonance imaging We conclude that hypo- and hyper- 10. Black KJ, Snyder AZ, Koller JM, Gado correlations. Diabetes Care 4:922–925, glycemia are associated with differences MH, Perlmutter JS: Template images for 1991 in regional grey and white matter volumes nonhuman primate neuroimaging. 1. Ba- 24. Auer RN, Hugh J, Cosgrove E, Curry B: in the brain in youth with type 1 diabetes. boon. Neuroimage 14:736–743, 2001 Neuropathological findings in three cases Longitudinal follow-up of well-character- 11. Mechelli A, Price CJ, Friston KJ, Ash- of profound hypoglycemia. Clin Neuro- ized samples such as ours is necessary to burner J: Voxel-based morphometry of path 8:63–68, 1989 determine the course of brain changes the human brain: methods and applica- 25. Tupola S, Salonen I, Hannonen R, Verho tions. Current Medical Imaging Reviews with age and further exposure to glycemic S, Saar P, Riikonen R: Comparison of re- 1:105–113, 2005 gional cerebral perfusion, EEG and cogni- extremes. Ultimately, an understanding 12. Salmond CH, Ashburner J, Vargha- tive functions in type 1 diabetic children of the implications of these findings for Khadem F, Connelly A, Gadian DG, Fris- with and without severe hypoglycaemia. optimal cognitive and academic function ton KJ: Distributional assumptions in Eur J Pediatr 163:335–336, 2004 must be obtained to place these observa- voxel-based morphometry. Neuroimage 26. Cavanna AE, Trimble MR: The precune- tions in proper clinical context. 17:1027–1030, 2002 us: a review of its functional anatomy and

2336 DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 Perantie and Associates

behavioural correlates. Brain 129:564– copy in children with poorly controlled 34. Worsley KJ, Andermann M, Koulis T, 583, 2006 type 1 diabetes mellitus. Neuroradiology MacDonald D, Evans AC: Detecting 27. Wessels AM, Simsek S, Remijnse PL, Velt- 47:562–565, 2005 changes in nonisotropic images. Hum man DJ, Biessels GJ, Barkhof F, Scheltens 31. Barnea-Goraly N, Menon V, Eckert M, Brain Mapp 8:98–101, 1999 P, Snoek FJ, Heine RJ, Rombouts SA: Tamm L, Bammer R, Karchemskiy A, 35. Yamada KA, Rensing N, Izumi Y, De Er- Voxel-based morphometry demonstrates Dant CC, Reiss AL: White matter devel- ausquin GA, Gazit V, Dorsey DA, Herrera reduced grey matter density on brain MRI opment during childhood and adoles- DG: Repetitive hypoglycemia in young in patients with diabetic retinopathy. Dia- cence: a cross-sectional diffusion tensor rats impairs hippocampal long-term po- betologia 49:2474–2480, 2006 imaging study. Cereb Cortex 15:1848– tentiation. Pediatr Res 55:372–379, 2004 28. DeVolder AG, Gofﬁnet AM, Bol A, Michel 1854, 2005 36. Malone JI, Hanna SK, Saporta S: Hyper- C, de Barsy T, Laterre C: Brain glucose 32. Schmithorst VJ, Wilke M, Dardzinski BJ, glycemic brain injury in the rat. Brain Res metabolism in postanoxic syndrome: Holland SK: Cognitive functions correlate 1076:9–15, 2006 positron emission tomographic study. with white matter architecture in a nor- 37. Deary IJ, Langan SJ, Graham KS, Hepburn Arch Neurol 47:197–204, 1990 mal pediatric population: a diffusion ten- DA, Frier BM: Recurrent severe hypogly- 29. Shaw P, Greenstein D, Lerch J, Clasen L, sor MRI study. Hum Brain Mapp 26:139– cemia, intelligence, and speed of informa- Lenroot R, Gogtay N, Evans A, Rapoport 147, 2005 tion processing. Intelligence 16:337–359, J, Giedd J: Intellectual ability and cortical 33. Ferguson SC, Blane A, Wardlaw J, Frier 1992 development in children and adolescents. BM, Perros P, McCrimmon RJ, Deary IJ: 38. Gold AE, MacLeod KM, Frier BM: Fre- Nature 440:676–679, 2006 Inﬂuence of an early-onset age of type 1 quency of severe hypoglycemia in pa- 30. Sarac K, Akinci A, Alkan A, Aslan M, Bay- diabetes on cerebral structure and cogni- tients with type 1 diabetes with impaired sal T, Ozcan C: Brain metabolite changes tive function. Diabetes Care 28:1431– awareness of hypoglycemia. Diabetes Care on proton magnetic resonance spectros- 1437, 2005 17:697–703, 1994

DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 2337